[llvm] 673c324 - [VectorCombine] foldInsExtVectorToShuffle - canonicalize new shuffle(undef,x) -> shuffle(x,undef).

[Simon Pilgrim via llvm-commits]

Author: Simon Pilgrim
Date: 2024-12-11T16:09:13Z
New Revision: 673c324ae3653cf62d67c5acbee1126e9eb6843e

URL: https://github.com/llvm/llvm-project/commit/673c324ae3653cf62d67c5acbee1126e9eb6843e
DIFF: https://github.com/llvm/llvm-project/commit/673c324ae3653cf62d67c5acbee1126e9eb6843e.diff

LOG: [VectorCombine] foldInsExtVectorToShuffle - canonicalize new shuffle(undef,x) -> shuffle(x,undef).

foldInsExtVectorToShuffle is likely to be inserting into an undef value, so make sure we've canonicalized this to the RHS in the folded shuffle to help further VectorCombine folds.

Minor tweak to help #34072

Added: 
    

Modified: 
    llvm/lib/Transforms/Vectorize/VectorCombine.cpp
    llvm/test/Transforms/VectorCombine/X86/extract-binop-inseltpoison.ll
    llvm/test/Transforms/VectorCombine/X86/extract-binop.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index 033900de55278c..9003642f1f93b2 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -2848,6 +2848,12 @@ bool VectorCombine::foldInsExtVectorToShuffle(Instruction &I) {
   if (OldCost < NewCost)
     return false;
 
+  // Canonicalize undef param to RHS to help further folds.
+  if (isa<UndefValue>(DstVec) && !isa<UndefValue>(SrcVec)) {
+    ShuffleVectorInst::commuteShuffleMask(Mask, NumElts);
+    std::swap(DstVec, SrcVec);
+  }
+
   Value *Shuf = Builder.CreateShuffleVector(DstVec, SrcVec, Mask);
   replaceValue(I, *Shuf);
 

diff  --git a/llvm/test/Transforms/VectorCombine/X86/extract-binop-inseltpoison.ll b/llvm/test/Transforms/VectorCombine/X86/extract-binop-inseltpoison.ll
index 3cf2940e86808e..80a21cef4e38ad 100644
--- a/llvm/test/Transforms/VectorCombine/X86/extract-binop-inseltpoison.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/extract-binop-inseltpoison.ll
@@ -474,8 +474,7 @@ define <4 x float> @PR34724(<4 x float> %a, <4 x float> %b) {
 ; SSE-NEXT:    [[TMP2:%.*]] = fadd <4 x float> [[B]], [[SHIFT1]]
 ; SSE-NEXT:    [[SHIFT2:%.*]] = shufflevector <4 x float> [[B]], <4 x float> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 2>
 ; SSE-NEXT:    [[TMP3:%.*]] = fadd <4 x float> [[SHIFT2]], [[B]]
-; SSE-NEXT:    [[V1:%.*]] = shufflevector <4 x float> poison, <4 x float> [[TMP1]], <4 x i32> <i32 0, i32 6, i32 2, i32 3>
-; SSE-NEXT:    [[V2:%.*]] = shufflevector <4 x float> [[V1]], <4 x float> [[TMP2]], <4 x i32> <i32 0, i32 1, i32 4, i32 3>
+; SSE-NEXT:    [[V2:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> [[TMP2]], <4 x i32> <i32 poison, i32 2, i32 4, i32 poison>
 ; SSE-NEXT:    [[V3:%.*]] = shufflevector <4 x float> [[V2]], <4 x float> [[TMP3]], <4 x i32> <i32 0, i32 1, i32 2, i32 7>
 ; SSE-NEXT:    ret <4 x float> [[V3]]
 ;
@@ -489,7 +488,7 @@ define <4 x float> @PR34724(<4 x float> %a, <4 x float> %b) {
 ; AVX-NEXT:    [[B01:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
 ; AVX-NEXT:    [[SHIFT2:%.*]] = shufflevector <4 x float> [[B]], <4 x float> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 2>
 ; AVX-NEXT:    [[TMP3:%.*]] = fadd <4 x float> [[SHIFT2]], [[B]]
-; AVX-NEXT:    [[V1:%.*]] = shufflevector <4 x float> poison, <4 x float> [[TMP1]], <4 x i32> <i32 0, i32 6, i32 2, i32 3>
+; AVX-NEXT:    [[V1:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> poison, <4 x i32> <i32 4, i32 2, i32 6, i32 7>
 ; AVX-NEXT:    [[V2:%.*]] = insertelement <4 x float> [[V1]], float [[B01]], i32 2
 ; AVX-NEXT:    [[V3:%.*]] = shufflevector <4 x float> [[V2]], <4 x float> [[TMP3]], <4 x i32> <i32 0, i32 1, i32 2, i32 7>
 ; AVX-NEXT:    ret <4 x float> [[V3]]

diff  --git a/llvm/test/Transforms/VectorCombine/X86/extract-binop.ll b/llvm/test/Transforms/VectorCombine/X86/extract-binop.ll
index a08506840572c5..307fbf711cdc2d 100644
--- a/llvm/test/Transforms/VectorCombine/X86/extract-binop.ll
+++ b/llvm/test/Transforms/VectorCombine/X86/extract-binop.ll
@@ -474,7 +474,7 @@ define <4 x float> @PR34724(<4 x float> %a, <4 x float> %b) {
 ; SSE-NEXT:    [[TMP2:%.*]] = fadd <4 x float> [[B]], [[SHIFT1]]
 ; SSE-NEXT:    [[SHIFT2:%.*]] = shufflevector <4 x float> [[B]], <4 x float> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 2>
 ; SSE-NEXT:    [[TMP3:%.*]] = fadd <4 x float> [[SHIFT2]], [[B]]
-; SSE-NEXT:    [[V1:%.*]] = shufflevector <4 x float> undef, <4 x float> [[TMP1]], <4 x i32> <i32 0, i32 6, i32 2, i32 3>
+; SSE-NEXT:    [[V1:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> undef, <4 x i32> <i32 4, i32 2, i32 6, i32 7>
 ; SSE-NEXT:    [[V2:%.*]] = shufflevector <4 x float> [[V1]], <4 x float> [[TMP2]], <4 x i32> <i32 0, i32 1, i32 4, i32 3>
 ; SSE-NEXT:    [[V3:%.*]] = shufflevector <4 x float> [[V2]], <4 x float> [[TMP3]], <4 x i32> <i32 0, i32 1, i32 2, i32 7>
 ; SSE-NEXT:    ret <4 x float> [[V3]]
@@ -489,7 +489,7 @@ define <4 x float> @PR34724(<4 x float> %a, <4 x float> %b) {
 ; AVX-NEXT:    [[B01:%.*]] = extractelement <4 x float> [[TMP2]], i32 0
 ; AVX-NEXT:    [[SHIFT2:%.*]] = shufflevector <4 x float> [[B]], <4 x float> poison, <4 x i32> <i32 poison, i32 poison, i32 poison, i32 2>
 ; AVX-NEXT:    [[TMP3:%.*]] = fadd <4 x float> [[SHIFT2]], [[B]]
-; AVX-NEXT:    [[V1:%.*]] = shufflevector <4 x float> undef, <4 x float> [[TMP1]], <4 x i32> <i32 0, i32 6, i32 2, i32 3>
+; AVX-NEXT:    [[V1:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> undef, <4 x i32> <i32 4, i32 2, i32 6, i32 7>
 ; AVX-NEXT:    [[V2:%.*]] = insertelement <4 x float> [[V1]], float [[B01]], i32 2
 ; AVX-NEXT:    [[V3:%.*]] = shufflevector <4 x float> [[V2]], <4 x float> [[TMP3]], <4 x i32> <i32 0, i32 1, i32 2, i32 7>
 ; AVX-NEXT:    ret <4 x float> [[V3]]